Proceedings of the 19th World Congress

The International Federation of Automatic Control

Cape Town, South Africa. August 24-29, 2014

Feedback Linearization-Based Control Approach for Air System of a

Turbocharged SI Engine: Toward a Fuel-Optimal Strategy

AnhTu Nguyen, Michel Dambrine, Jimmy Lauber Michio Sugeno

LAMIH, UMR CNRS 8201, University of Valenciennes, France European Center for Soft Computing
E-mail: nguyen.trananhtu@gmail.com, {michel.dambrine, E-mail: michio.sugeno@gmail.com
jimmy.lauber}@univ-valenciennes.fr

Abstract: Combining turbocharging with downsizing has now become a key technique for automotive
engine to improve its performance as fuel economy and drivability. The technology potential is fully
exploited only with an efficient air path management system. In this context, the present work addresses
a novel control design based on feedback linearization for the turbocharged air system of a SI engine.
Two control strategies have been investigated for this complex system: drivability optimization and fuel
reduction. The effectiveness of the proposed approach is evaluated via an advanced engine simulator.
Keywords: Turbocharged SI engine, feedback linearization, MIMO approach, fuel-optimal strategy.

1. Introduction
The air system control issue of turbocharged engine is
known as a very interesting problem in automotive industry.
Over the years, many control approaches have been proposed
in the literature. However, up to now, it is still an active
research subject in industry. The difficulties when dealing
with this system are mainly due to the following facts. First,
this MIMO system has a complex dynamics with several
involved nonlinearities. Second, it is not easy to take into
account the fuel-optimal strategy [1] in the control design
when considering the whole system.
In the literature, most of works propose a decentralized
control approach, i.e. one controller for each air actuator. In
addition, linearization around operating points is often
needed [2,3]. These controllers have some unavoidable
drawbacks due to the application of linear control techniques
to nonlinear systems. First, the tradeoff between performance
and robustness throughout the wide engine operating range is
difficult to achieve. Second, the calibration effort is very
expensive. In [4], the authors proposed an interesting
nonlinear approach for controlling the turbocharger of a SI
engine which may significantly reduce the calibration task.
However, this SISO approach allows only the wastegate
control of the turbocharged engine air system. In addition, it
needs some costly model simplification task. In [5], we have
proposed a novel control strategy based on a switching
Takagi-Sugeno model [6] which can get rid of the
aforementioned difficulties. Although this powerful nonlinear
controller provides satisfying closed-loop performance, it
may look complex from an industrial point of view. In this
paper, a second approach based on feedback linearization for
the turbocharged air system which is much simpler (in the
sense of control design and implementation) and can achieve
practically a similar level of performance as the previous one.
To the best of our knowledge, this is the second nonlinear
MIMO controller that can guarantee the stability of the whole
closed-loop turbocharged air system while taking into
account the fuel-optimal strategy after [5,7] and the first
nonlinear controller which is directly based on the complete
nonlinear model of this system. Furthermore, the control
approach proposed in this work could also limit the costly
automotive sensors and/or observers/estimators design tasks
by estimating all variables needed for control design by their
static look-up-tables issued from the data of the test bench.
Despite its simplicity, the simulation results carried out on an
advanced simulator clearly show the effectiveness of the
proposed control approach.
The paper is organized as follows. Section 2 presents a
brief description on the air system of a turbocharged SI
engine. Some bases on feedback linearization technique are
recalled in Section 3. Section 4 is devoted to nonlinear
control design and shows how to derive the Fuel-optimal
controller from the proposed MIMO controller. Some
simulation results are reported in Section 5. Finally, a
concluding remark is provided in Section 6.
Figure 1. Schematic of a turbocharged SI engine
2. Turbocharged SI Engine Air System:
a Brief Description
The studied air system is depicted in Figure 1. Hereafter,
some main equations governing its behavior are recalled,
please refer to [8,9] for more details.

978-3-902823-62-5/2014 © IFAC 2994

19th IFAC World Congress
Cape Town, South Africa. August 24-29, 2014

The following notations will be used in this paper: 2γ

Variables Description Unit
where Φ ( Π ) ≜
γ −1
( )
Π 2 γ − Π (γ +1) γ , Π*thr ≜ max ( Π thr , φ ) ,
Pboost Boost pressure Pa γ
 2  γ −1
Π ≜ max ( Π wg , φ ) and φ ≜ 
Pman Intake pressure Pa *
 .
 γ +1 
wg
Pexh Exhaust pressure Pa
Pamb Atmospheric pressure Pa In (3), the two control inputs uthr ≜ Sthr and uwg ≜ S wg are
Π thr ≜ Pman Pboost Throttle pressure ratio -- respectively the effective actuator opening sections of the
Π wg ≜ Pdt Pexh Wastegate pressure ratio -- throttle and of the wastegate. The in-cylinder air flow in (1) is
computed as [8]:
Π comp ≜ Pboost Pamb Compressor pressure ratio --
PmanVcyl N e
Π turb ≜ Pdt Pexh Turbine pressure ratio -- Dcyl = ηvol (4)
RTman 30
Tamb Atmospheric temperature °K
Tman where ηvol = LUTηvol ( N e , Pman ) is given by a look-up-table
Intake manifold temperature °K
Texh Exhaust manifold temperature °K (LUT). Assume that the SI engine operates at stoichiometry;
Dthr
the fuel injected flow can be then deduced as follows:
Throttle mass air flow kg/s
1
Dwg Wastegate mass air flow kg/s D fuel = Dcyl (5)
λs
Dcomp Compressor mass air flow kg/s
Dturb Turbine mass air flow kg/s 2.3. Turbocharger Modeling
Dcyl Cylinder mass air flow kg/s The turbocharger dynamics is modeled as:
D fuel Fuel injected flow kg/s d 1 2 
 J tc N tc  = Pturb − Pcomp (6)
Vexh Exhaust manifold volume m3 dt  2 
Vman Intake manifold volume m3 where J tc is the inertia of the turbocharger. The power
Vcyl Cylinder volume m3 consumed by the compressor is:
γ −1
Ne Engine speed rpm 1  γ 
Pcomp = Dcomp C pTamb  Π comp − 1 (7)
Pcomp Compressor power W ηcomp  
Π comp = LUTΠ comp ( N tc , Dcomp )
Pturb Turbine power W where the quantities and
ηcomp Compressor isentropic efficiency --
ηturb Turbine isentropic efficiency --
ηcomp = LUTη comp
(N tc , Dcomp ) are mapped.
ηvol Engine volumetric efficiency -- Similarly, the power delivered by the turbine can be written:
1-γ
λs Stoichiometric air/fuel ratio --  
Pturb = Dturb C pTexhηturb  1 − Π turb
γ
 (8)
γ Isentropic coefficient -- 
 
R Ideal gas constant J/kg/°K
Cp where the quantities ηturb = LUTηturb ( N tc , Π turb ) and
Specific heats at constant pressure J/kg/°K
Dturb = LUTDturb ( N tc , Π turb ) are mapped.
2.1. Intake and Exhaust Pressure Dynamics
Note that the turbocharger variables (air flow and speed)
Here, it is assumed that the volumes of the intake and are corrected to consider the variations of the thermodynamic
exhaust manifolds are fixed, and that their thermodynamic conditions in the upstream (resp. downstream) of the
conditions are homogeneous. In addition, the temperature compressor (resp. turbine) [10].
dynamics are supposed to be negligible. Then, the isothermal
filling-emptying model [8] is used to derive the intake and 2.5. Complete Model for Control Design
exhaust pressure dynamics: By grouping equations (1), (2) and (6), the following
Pɺman = man ( Dthr − Dcyl )
RT dynamical system is obtained:
(1)
Vman  dPman
 = K man ( Dthr − Dcyl )
Pɺexh = exh ( Dcyl + D fuel − Dturb − Dwg )
RT
(2)  dt
 dPexh
Vexh
 = K exh ( K fuel Dcyl − Dturb − Dwg ) (9)
Next, the mass air flow calculations will be detailed.  dt
d 1 2 
2.2. Air Flows Calculations   J tc N tc  = Pturb − Pcomp
 dt  2 
The actuator valves are modeled with the standard
The model (9) accurately describes the three main dynamics
equations of compressible gas flow through a nozzle [9]:
governing the turbocharged air system of a SI engine. This
Dthr = Φ ( Π *thr ) boost uthr ; Dwg = Φ ( Π *wg ) exh u wg (3)
P P model will be directly used to design the nonlinear MIMO
RTman RTexh controller in Section 4.

2995
19th IFAC World Congress
Cape Town, South Africa. August 24-29, 2014
3. Nonlinear Control System Design
Hereafter, only some bases on input-output linearization
for control-affine MIMO nonlinear systems will be reminded,
please refer to [11] for more details.
3.1. Input-Output Linearization for MIMO system
Consider the following MIMO nonlinear system:
 m
 ɺ
x = f ( x ) + ∑ gi ( x ) ui
 i =1
 y = h x ≜  h x ,… , h x  T
 ( )  1( ) m ( )
where x ∈ ℝ n , u ∈ ℝ m and y ∈ ℝ m are respectively the
system state, control input, and output vectors. The vector
functions f ( x ) , g ( x ) and h ( x ) are assumed to be
sufficiently smooth in a domain D ⊂ ℝ . The feedback n for the air system presented in Section 2.
linearization law of the system (10) is given as:
−1
 Lg1 Lρf1 −1h1 ( x ) … Lgm Lρf1 −1h1 ( x )    v1   Lρf1 h1 ( x )  
    
u= … … …   ⋮  − 
 L Lρm −1h ( x ) … L Lρm −1h ( x )   v   Lρm h ( x )   (11)
 g1 f m gm f m
   m   f
≜ J −1 ( x ) ( v − l ( x ) )
where [ ρ1 … ρ m ] is the relative degree vector; Lρfi hi ( x ) and
T
Lgi Lρf i −1hi ( x ) are the Lie derivatives of the scalar functions
hi ( x ) , i = 1,… , m . Note that the control law (11) is well
defined in the domain D ⊂ ℝ n if the decoupling matrix
J ( x ) is non-singular at every point x0 ∈ D ⊂ ℝ n . The new
manipulated input vector v can be designed with any linear
control technique. The relative degree of the system (10) in
this case is defined as:
m It is worth noting from (14) that all the inputs of these LUTs
ρ = ∑ ρk
k =1
3.2. Normal Form and Internal Dynamics Analysis
The system relative degree ρ plays an important role in
feedback linearization control technique. Indeed, according to
its value, three following cases are considered.
• Case 1: If ρ = n , then the nonlinear system (10) is fully
feedback linearizable.
• Case 2: If ρ < n , then the nonlinear system (10) is
partially feedback linearizable. In this case, there are some
internal dynamics of order ( n − ρ ) . In tracking control, it
should be guaranteed that these dynamics are well behaved,
i.e. stable or bounded in some sense.
• Case 3: If ρ does not exist on the domain D , then the
input-output linearization technique is not directly applicable
with this technique.
The linearized system for the two first cases can be
represented under the following normal form [11]:
ξɺ = Aξ + Bv
  yman ≜ Pman
 y = Cξ (13)
ωɺ = f ω , ξ , v
0( )
  yexh ≜ Pexh
where ξ and ω are respectively ρ -dimensional and
( n − ρ ) -dimensional state vector which are obtained with a
suitable change of coordinates z ≜ T ( x ) ≜ [ξ , ω ] ; the triplet
T
( A, B, C ) is in Brunovsky block canonical form; the last
equation in (13) characterizes the internal dynamics [11]. It is
worth noting that if the system ωɺ = f 0 (ω , ξ , v ) is input-to-
state stable, then the origin of system (13) is globally
(10) asymptotically stable [12] if v is a stabilizing control law for
the first subsystem.
4. Application: Control of the Air System of a
Turbocharged SI Engine
In this section, a MIMO nonlinear controller is designed
It is first very important to highlight the following fact.
For almost all controllers existing in actual literature, not
only available measures of engine intake side (i.e. Pman , Tman ,

⋮  Pboost , N e ) but also several other signals coming from the
m  exhaust side are needed for controller implementation. These
latter signals, i.e. Pexh , Texh , Pdt , N tc are not measured in
series production vehicles and usually assumed to be given
by some observers. To get rid of this assumption, in this
work, these variables will be estimated by their static LUTs
issued from the data measured in steady-state conditions in
the test bench. As a consequence, the number of sensors
or/and complex observers/estimators could be limited.
Concretely, the following LUTs will be constructed:
 Pexh = LUTP ( N e , Pman ) ; Texh = LUTT ( N e , Dcyl )
 exh exh
 (14)
 Pdt = LUTPdt ( N e , Dcyl ) ; N tc = LUTNtc ( Π comp , Dcomp )
(12) Pexh , Texh , Pdt , N tc can be measured/computed with
available sensors. In fact, the estimations in (14) are
reasonable since SI engines operate at stoichiometric
conditions which implies that all exhaust variables are highly
correlated to the in-cylinder air mass flow (or intake manifold
pressure) [9].
4.1. MIMO Controller Design
For control design purpose, besides the output of interest
yman ≜ Pman , a second virtual one is introduced yexh ≜ Pexh .
Note that, by means of LUT in (14), we can impose that
Pexh, ref = LUTPexh ( N e , Pman , ref ) and, then, if Pexh converges to
Pexh , ref , it implicitly makes Pman converge to Pman , ref . As a
consequent, both outputs Pman and Pexh are used to control
the intake pressure. Now, let us first consider the two
pressure dynamics in (9) which are rewritten as:
 Pɺman = K man ( Dthr − Dcyl ) ≜ f thr + gthr uthr

 Pɺexh = K exh ( K fuel Dcyl − Dturb − Dwg ) ≜ f wg + g wg uwg
 (15)

where
2996
19th IFAC World Congress
Cape Town, South Africa. August 24-29, 2014

 It follows from (22) and (24) that:

; g thr ≜ K man boost Φ ( Π*thr ) ;
RTman P
 K man ≜ V
 man RTman d
dt
( Ntc2 ) = − Kturb Dwg − Kcomp Dcomp + Kturb K fuel Dcyl − KKturb vexh (25)
 V N exh

 K cyl ≜ ηvol cyl e Pman ; f thr ≜ − K man Dcyl ; Note that P ≜ [ Pman , Pexh ] can be seen as the input vector of
T

 Vman 30
 (16) system (25) and it follows from (21) that:
 K ≜ 1 + 1  ; g ≜ − K Pexh
Φ ( Π wg ) ;
*
 fuel  λ 
2
K 
( Kturb K fuel K cyl ) +  Kturb K( 2,2)  P − K comp Dcomp (26)
wg exh
Nɺ tc2 <
2
  s  RTexh
 RTexh  exh 
 K exh ≜ ; f wg ≜ K exh ( K fuel Dcyl − Dturb ) .
 Vexh ≜ α ( ⋅) P − β ( N tc2 )
The time derivatives of the outputs of system (15) are given: Since α ( ⋅) is bounded and the function β ( ⋅) is of class K ∞
ɺ
 yɺ man = Pman = f thr + gthr uthr = vman (its curve form is not depicted here due to lack of space).
 ɺ
(17)
 yɺ exh = Pexh = f wg + g wg uwg = vexh Then, it can be deduced that the system (22) is always input-
to-state stable [13].
Remark that the two control inputs uthr , uwg appear
4.2. Fuel-Optimal Control Strategy
respectively in yɺ man , yɺ exh ; the signals vman and vexh are two
new manipulated inputs. By using integral structure for Until now, we have designed in this work a so-called
tracking control purpose, the linearized system Conventional MIMO controller with two inputs: throttle,
 yɺ man = vman wastegate and two outputs: intake pressure, exhaust pressure.
 However, this controller is not optimal in the sense of energy
 yɺ exh = vexh (18) losses minimization. Indeed, it is known that the wastegate
 xɺ = y
man , ref − yman
should be opened as much as possible at a given operating
 int
point to minimize the pumping losses [1]. This concept
is straightforwardly derived from the system (15) with the
generally leads to the following strategy: in low load zone,
following feedback linearization control laws:
only the throttle is used to track the intake pressure, the
 f thr 1
uthr = − g + g vman
wastegate is widely open and in high load zone, the wastegate
 thr thr is activated to control the pressure while the throttle is widely
 f
(19) open. Although this strategy offers many advantages
u = − wg + 1 v concerning the fuel economy benefits and also the control
 wg g wg g wg
exh
 design simplification (only one actuator is controlled at a
Let us define x ≜ [ yman , yexh , xint ] and v ≜ [ vman , vexh ] . Then, time), it may overcharge the wastegate in some cases since
T T

the torque response is slower for this actuator than for the
the linearized system (18) is rewritten as: throttle. To overcome this eventual difficulty while taking
 0 0 0 1 0 0 fully into account the fuel-optimal strategy, we propose here
     
xɺ =  0 0 0  x +  0 1  v +  0  yman , ref (20) a so-called Fuel-optimal controller. This novel controller is
 −1 0 0  0 0 1 directly derived from the Conventional MIMO controller and
     
they both have the same auxiliary control law given in (21).
In this work, the state feedback law v = − K ξ is designed It can be deduced from the second equation of (15) that:
through a simple pole-placement approach. The following vexh D Dwg
control gain will be implemented in the simulator: Dcyl = + turb + (27)
K exh K fuel K fuel K fuel
140 0 −2400 
v = − Kx = −  x (21)
0 
With this new expression of the in-cylinder mass air flow, the
 0 50 intake pressure dynamics can be also rewritten as:
Note that the tracking problem has been solved accounting  vexh D D 
only for a part of the closed-loop system (9). The stability Pɺman = K man  Dthr − − turb − wg  = vman
 K exh K fuel K fuel K fuel 
analysis of the following internal dynamics N tc2 is necessary: 
or (28)
d
( Ntc2 ) = Kturb Dturb − K comp Dcomp (22) K K K
dt Pɺman = gthr uthr − man
vexh − man Dturb + man
g wg uwg
where K exh K fuel K fuel K exh K fuel
 2  1-γ
 The novel Fuel-optimal controller is directly derived from
 K turb ≜ C pTexhηturb  1 − Π turb
γ
 expression (28). To this end, the engine operating range is
 J tc 
  divided into three zones according to two predefined intake
 γ -1
(23)
 2 1  γ  pressure thresholds Pman,1 and Pman ,2 . Each operating zone
K ≜
 comp J p amb η C T  Π comp − 1 
  has its own actuator scheduling strategy as described below:
 tc comp  
Furthermore, from (15) and (17), one gets: • Zone 1 (low load zone Pman ≤ Pman ,1 ): The wastegate is
v widely open and the throttle is solely used to track the intake
Dturb = K fuel Dcyl − Dwg − exh (24) pressure reference. Let S wg ,max be the maximal opening
K exh

2997
19th IFAC World Congress
Cape Town, South Africa. August 24-29, 2014

section of the wastegate. The implemented actuator control 5. Simulation Results and Analysis
laws are in this case:

Hereafter, a series of trials are performed on an advanced
1  K man K K man 
uthr = 

vexh + man Dturb − g wg S wg ,max + vman 

engine simulator designed under commercial AMESim
 g thr  K exh K fuel K fuel K exh K fuel  (29) platform to show the effectiveness of the proposed approach

uwg = S wg ,max for both cases: Conventional MIMO controller and Fuel-
• Zone 2 (middle load zone Pman ,1 < Pman < Pman,2 ): Both optimal controller. For the sake of clarity, the two commands
(throttle, wastegate) are normalized. The control inputs
throttle and wastegate are simultaneously used to control the constraints become: 0 ≤ uthr , uwg ≤ 100% . When uthr = 100%
intake pressure. In this case, the implemented actuator control
laws are exactly the feedback linearization laws in (19), (resp. uwg = 0% ), it means that the throttle (resp. wastegate)
which are recalled here: is fully open. On the reverse, when uthr = 0% (resp.
f 1 f wg 1
uthr = − thr + vman ; uwg = − + vexh (30) uwg = 100% ), the throttle (resp. wastegate) is fully closed. In
g thr g thr g wg g wg
the following simulations at N e = 2000 rpm , the pressure
• Zone 3 (high load zone Pman ≥ Pman,2 ): The throttle is fully
thresholds are chosen as Pman ,1 = 0.9 bar and Pman ,2 = 1.2 bar.
opened and only the wastegate is activated to control the
intake pressure which is approximated by the boost pressure Figure 2 and Figure 3 respectively represent the tracking
performance and the corresponding actuator responses for
Pboost . The implemented actuator control laws are:
Conventional MIMO controller and Fuel-optimal controller.
uthr = Sthr ,max Several comments can be reported as follows. First,

 K exh K fuel  K man K  (31) Conventional MIMO controller simultaneously uses both
uwg =  vexh + man Dturb − gthr Sthr ,max + vman 
 actuators to track the intake pressure while these actuators are
 K man gthr  K exh K fuel K fuel  "optimally" scheduled by the strategy described in Subsection
where Sthr ,max is the maximal opening section of the throttle. 4.2 with Fuel-optimal controller. Second, the wastegate is
Several remarks can be made for this actuator scheduling opened very little with Conventional MIMO controller so that
strategy. First, the new input vector v is kept to be the same the boost potential of the turbocharger can be fully exploited.
for all three zones. Second, when the wastegate (resp. Hence, the closed-loop time response with this controller is
throttle) is saturated in Zone 1 (resp. Zone 3), the exhaust faster than the one of Fuel-optimal controller in middle and
(resp. intake) pressure dynamics is input-to-state stable with high load zones. Third, although Conventional MIMO
respect to the intake (resp. exhaust) pressure (the proof is controller can be used to improve the torque response
omitted due to lack of space). Third, it can be concluded from (drivability), this controller is not optimal in terms of fuel
the above remarks that if the intake pressure tracking is consumption compared with Fuel-optimal controller as
guaranteed, then all other variables of the air system will be shown in Figure 4. The pumping losses with Fuel-optimal
well behaved in spite of the fact that the engine operating controller are almost lower than the ones with Conventional
range is divided into three zones. Fourth, the model-based MIMO controller at every time.
Fuel-optimal controller is based on a "dummy" switching Figure 5 shows the comparison, in the case of Fuel-
strategy because no switching model has been used in this optimal controller, between the exhaust pressure given by the
simulator AMESim and its static LUT which is used to
approach. Fifth, the pressure threshold values Pman,1 and
compute the controller. Note that although some higher
Pman ,2 separating the three zones are "freely" chosen thanks dynamics are missing and there are some slight static errors,
to the propriety of the above fourth remark. However, they the tracking performance of both controllers is very
are usually very close for engine efficiency benefits. satisfying.
2
It is worth noting that Fuel-optimal controller is different
Intake Pressures (bar)

from other existing nonlinear approaches in the literature. As 1.5

the approach proposed in [5], this novel controller is a MIMO

1
nonlinear controller which can guarantee the closed-loop
stability of the whole turbocharged engine air system. 0.5

However, the novel Fuel-optimal controller is much simpler

0
and the middle-load zone (Zone 2) is very easily introduced 0 2 4 6 8 10 12 14 16 18 20
Time (s)
to improve the torque response at high load while
maintaining the maximum possible the advantage of fuel- 100

optimal concept in [1]. Compared with the approach in [4] 80 Wastegate

Actuators (%)

Throttle
which is also based on feedback linearization, our approach 60
does not need any model simplification, e.g. neglecting
40
pressure dynamics with respect to turbocharger dynamics
20
according to singular perturbation theory and then
approximating the turbocharger square speed as a linear 0
0 2 4 6 8 10 12 14 16 18 20
function of compressor pressure ratio. Moreover, in [4], the Time (s)

authors only proposed the control for the wastegate and this Figure 2. Pressure tracking performance and corresponding
approach cannot take into account the mid-load zone. actuator responses with Conventional MIMO controller

2998
19th IFAC World Congress
Cape Town, South Africa. August 24-29, 2014

2
Handling the model uncertainties is known as one of
Intake Pressures (bar)

1.5
Pman reference
Pman
major drawbacks of feedback linearization based controllers.
The study on robust design will be investigated in future
1
research.
0.5
Acknowledgments
0
0 2 4 6 8 10 12 14 16 18 20
This research was supported by the International Campus
Low load with Middle load with High load with Middle load with Low load with
THR only THR and WG WG only THR and WG THR only on Safety and Intermodality in Transportation, the Nord-Pas-
100
de-Calais Region, the European Community, the Regional
80 Delegation for Research an Technology, the Ministry of
Actuators (%)

Wastegate
60 Throttle Higher Education and Research, and the French National
40
Center for Scientific Research. The authors gratefully
acknowledge the VALEO Company for their supports in the
20
framework of the SURAL-HY project. We would also like to
0
0 2 4 6 8 10 12 14 16 18 20 thank Prof. Marie-Thierry Guerra, Director of LAMIH for his
Time (s) valuable supports in the collaboration framework with Prof.
Figure 3. Pressure tracking performance and corresponding Michio Sugeno, Emeritus Researcher from European Centre
actuator responses with Fuel-optimal controller for Soft Computing, Spain.
7

6
References
Energy Lost to Pumping (%)

Fuel-optimal controller
5 Conventional MIMO controller [1] L. Eriksson, S. Frei, C. Onder, and L. Guzzella, "Control and
4 Optimization of Turbocharged SI Engines," in 15th IFAC
3
World Congress, Barcelona, Spain, 2002.
2
[2] A. Karnik, J. Buckland, and J. Freudenberg, "Electronic
Throttle and Wastegate Control for Turbocharged Gasoline
1
Engines," in American Control Conference, Portland, 2005, pp.
0
0 2 4 6 8 10 12 14 16 18 20
4434-4439.
Time (s)
[3] L. Daubler, C. Bessai, and O. Predelli, "Tuning Strategies for
Figure 4. Comparison of engine pumping losses between Online-Adaptive PI Controller," Oil & Gas Science and
Conventional MIMO controller and Fuel-optimal controller Technology - Rev. IFP, vol. 62, no. 4, pp. 493-500, 2007.
1.8
[4] P. Moulin and J. Chauvin, "Modeling and Control of the Air
1.7
Pexh
System of a Turbocharged Gasoline Engine," Control
1.6 Pexh LUT Engineering Practice, vol. 19, no. 3, pp. 287-297, 2011.
Exhasut Pressure (bar)

1.5 [5] A.T. Nguyen, J. Lauber, and M. Dambrine, "Robust H∞ Control

1.4 Design for Switching Uncertain System: Application for
1.3 Turbocharged Gasoline Air System Control," in 51st
1.2
Conference on Decision and Control, Maui, Hawaii, USA,
2012, pp. 4265-4270.
1.1

1
[6] T. Takagi and M. Sugeno, "Fuzzy Identification of Systems and
Its Applications to Modeling and Control," Trans. on Systems,
0.9
0 2 4 6 8 10 12 14 16 18 20 Man and Cybernetics, vol. 15, no. 1, pp. 116-132, 1985.
Time (s)

Figure 5. Comparison between the exhaust pressure given by [7] AT. Nguyen, J. Lauber, and M. Dambrine, "Switching Fuzzy
Control of the Air System of a Turbocharged SI Engine," in
the simulator AMESim and its static LUT model IEEE International Conference on Fuzzy Systems, Brisbane,
Australia, 2012, pp. 893-899.
6. Conclusions
[8] L. Guzzella and C. Onder, Introduction to Modeling and
In this paper, a novel approach has been proposed to Control of Internal Combustion Engine Systems. Springer,
control the turbocharged air system of a SI engine. Several 2004.
advantages of this approach can be summarized as follows. [9] L. Eriksson, "Modeling and Control of Turbocharged SI and DI
First, the second virtual output yexh ≜ Pexh is introduced by Engines," Oil & Gas Science and Technologies, vol. 62, no. 4,
pp. 523-538, 2007.
means of LUT and this fact drastically simplifies the control
[10] P. Moraal and I. Kolmanovsky, "Turbocharger Modeling for
design task. Second, the resulting nonlinear control law is
Automotive Control Applications," SAE Technical Paper
very easily implementable. Third, offline engine data of the Series, vol. no. 1999-01-0908, 1999.
test bench is effectively reused and exploited for engine
[11] A. Isidori, Nonlinear Control Systems. Springer Verlag, 1989.
control development so that the number of sensors and/or
observers/estimators could be limited. Despite its simplicity, [12] H.K. Khalil, Nonlinear Systems, 3rd ed.: Prentice Hall, 2002.
the proposed approach performs very promising results for [13] E. Sontag and Y. Wang, "On Characterizations of the Input to
both control strategies, i.e. to improve the drivability with State Stability Property," Systems and Control Letters, vol. 24,
Conventional MIMO controller or to optimize the fuel pp. 351-359, 1995.
consumption with Fuel-optimal controller.

2999